Anthropic cuts off OpenAI’s access to its Claude models

Anthropic has withdrawn the access from OpenAi to his Claude family of AI models, according to A report in Wired.

Sources told Wired that OpenAi Claude connected with internal tools with which the company could compare the performance of Claude with its own models in categories such as coding, writing and safety.

In a statement to WAN, the anthropic spokesperson said: “OpenAi’s own technical staff also used our coding tools prior to the launch of GPT-5”, which apparently is “a direct violation of our service conditions”. (The commercial conditions of Anthropic prohibit companies to use Claude to build competitive services.)

However, the company also said that the OpenAi would continue to give access to “for evaluations of benchmarking and safety.”

In the meantime, a self -declaration described, a spokesperson for OpenAi described his use as a “industrial stand” and added: “Although we respect the decision of Anthropic to cut our API access, it is disappointing that our API remains available to them.”

Anthropic managers had already shown resistance to access to competitors, with Chief Science Officer Jared Kaplan who previously justified the decision of the company to reduce Windsurf (a rumor openi acquisition -then gained by Cognition) by saying: “I think it would be strange for us to sell Claudude.”